Global methane emissions from rivers and streams

Freshwater ecosystems contribute nearly half of global CH₄ emissions to the atmosphere, but the role of rivers and streams remains unclear despite indications that their total emissions may rival sources such as biomass burning and rice cultivation. Running waters both produce CH₄ internally and receive substantial external CH₄ from adjacent soils and wetlands, implying controls that span land–water boundaries. However, past syntheses show extreme spatial and temporal variability (spanning seven orders of magnitude) and strong fine-scale controls, leading to global estimates with large uncertainty, unknown spatial patterns, and discrepancies between bottom-up inventories and top-down constraints. To address these gaps, the study leverages a large global database (GRiMeDB) to model and upscale CH₄ concentrations and emissions, identify primary drivers, and provide spatially and seasonally explicit global estimates.

Previous work shows a strong and consistent temperature dependence of aquatic methanogenesis and CH₄ emissions across freshwater systems, with high apparent activation energies. Yet for rivers and streams, existing global estimates relied on simple averaging of sparse and highly variable measurements, producing wide uncertainty and discrepancies with atmospheric (top-down) budgets. Additional complexities include multiple emission pathways (diffusion and ebullition), fine-scale heterogeneity in groundwater inputs and sediments, and hydrologic variability. Prior studies also highlight the importance of fluvial–wetland connectivity (for example, in the Congo), source limitation in high-elevation systems, and human impacts via impoundments, agriculture, and urban wastewater. These findings motivated a data-rich, model-based reassessment focused on cross-boundary drivers, geomorphology, and hydrology.

- Database: Used the Global River Methane database (GRiMeDB) comprising >24,000 CH₄ concentration observations and >8,000 CH₄ flux observations worldwide. - Modeling approach: Trained monthly random forest machine-learning models to predict CH₄ concentrations globally using spatial predictors spanning climate, edaphic/soil properties, land cover, human influence, and physical/geomorphological variables (for example, river slope, elevation, gas-transfer velocity, groundwater table depth, peatland cover, net primary production, soil respiration). Model performance across months explained a substantial fraction of variability (R² between log-transformed modeled versus withheld observations = 0.45–0.68). - Upscaling: Produced seasonally and spatially explicit maps of CH₄ concentrations and converted to diffusive emissions using gas-transfer velocity formulations, with a posteriori corrections that especially affect mountainous regions. Emissions were aggregated globally and by latitude bands. - Treatment of ice: Assumed ice/snow cover prevents riverine CH₄ emissions, acknowledging this is conservative given potential under-ice buildup and release during ice break-up. - Human-modified systems: Excluded the most highly modified systems (for example, direct point sources, ditches, canals, and strongly impacted reaches) from the predictive modeling due to lack of adequate spatial predictors; however, these settings were analyzed separately to document elevated concentrations. - Ebullition: Recognizing sparse ebullitive measurements and high variability, estimated ebullitive CH₄ emissions using observed linear relationships between diffusive and ebullitive fluxes (log-transformed data) from GRiMeDB and literature (Extended Data), yielding a global ebullition estimate and associated uncertainty. - Uncertainty: Reported 10th–90th percentile ranges using Monte Carlo simulations for diffusive and total emissions.

- Global magnitude: Total global CH₄ emissions from rivers and streams are 27.9 (16.7–39.7) Tg CH₄ per year, similar in magnitude to other freshwater systems. - Diffusive emissions: 13.4 (10.1–16.8) Tg CH₄ per year. - Ebullitive emissions: 14.5 (6.6–22.9) Tg CH₄ per year. - Spatial patterns: Highest concentrations and emissions occur in the tropics (Southeast Asia, Congo Basin, Pantanal and Amazon floodplains), but elevated values also occur in Arctic and boreal biomes (Fennoscandia, Alaska, Eastern Siberia). - Latitudinal contributions: Tropics (10° S–10° N) account for the largest share of emissions (37%). High-latitude Arctic and northern boreal regions (>50° N) contribute ~17%, comparable to temperate/subtropical 30–50° N (~15%). - Drivers: Emissions and concentrations are linked to edaphic and climate features promoting anoxia (high organic matter supply, water saturation, shallow groundwater tables, peatland cover, large soil organic carbon stocks), and regulated at reach scales by physical/geomorphic variables (river slope, elevation, gas-transfer velocity; negative effects on concentrations). - Temperature sensitivity: Apparent activation energy for riverine diffusive emissions is low (site-specific median Eₐ ≈ 0.14 eV; across dataset ≈ 0.17 eV), significantly lower than lakes/wetlands/rice paddies (~0.96 eV), indicating temperature is not a first-order global control for running waters. - Seasonality: Marked seasonal dynamics, especially at high latitudes due to ice cover, open-water season length, and hydrological connectivity between wet and dry seasons. - Human influence: Population density positively associated with concentrations. Elevated CH₄ observed downstream of point sources (wastewater), in ditches, urban canals, and reaches influenced by thermogenic inputs or dams. Reservoirs/impoundments (not included here) contribute substantially (~10% of freshwater CH₄ emissions) as lentic systems. - Model performance: Random forest models captured substantial variability (R² 0.45–0.68 for log-transformed concentrations) and reduced uncertainty relative to prior global estimates.

The study resolves key uncertainties in the contribution of rivers and streams to the global CH₄ budget, demonstrating that total emissions are globally important and comparable to other freshwater sources. Elevated emissions in both warm and cold regions reflect the primacy of landscape-scale CH₄ production and supply (edaphic, hydrologic connectivity, organic matter availability) rather than a simple thermal control. The open and advective nature of running waters integrates external CH₄ inputs from soils and wetlands and modulates temperature sensitivity via mixing, oxidation, and transit. Geomorphology and hydraulics further regulate in-stream concentrations and emissions by enhancing gas exchange and limiting gas accumulation. The findings imply that climate change will influence riverine CH₄ emissions primarily through indirect effects on terrestrial CH₄ generation, precipitation-driven connectivity, and loading of organic matter and nutrients, rather than only via warming of water temperatures. The demonstrated human influences suggest opportunities for mitigation by managing wastewater, nutrient and organic matter inputs, and hydrological modifications.

This work provides a spatially and seasonally explicit global estimate of riverine CH₄ emissions (27.9 Tg CH₄ per year), clarifies the dominant drivers across scales, and shows that emissions are governed more by land–water connectivity and catchment properties than by temperature alone. The results reduce uncertainty relative to prior estimates and are suitable for incorporation into global CH₄ budgets. Future research should: (1) expand measurements of ebullition in running waters to better constrain global totals; (2) develop process-based models of riverine CH₄ that explicitly represent landscape and hydrological controls (soil saturation, wetland connectivity, organic matter supply); (3) quantify impacts of changing ice cover, precipitation regimes, and permafrost thaw on CH₄ supply to streams; and (4) assess and mitigate human-enhanced emissions by targeting point sources, urban and agricultural drainage networks, and river network modifications.

- Model predictors are relatively coarse in space and time, applied to monthly aggregated concentrations; fine-scale heterogeneity (groundwater inputs, sediment properties, diel variability) is not captured and contributes to unexplained variance. - Assumption that ice/snow cover prevents emissions likely underestimates high-latitude winter/shoulder-season fluxes and pulse emissions during ice break-up. - Upscaled fluxes are sensitive to gas-transfer velocity corrections, especially in mountainous regions. - Paucity and high variability of ebullitive measurements from running waters required indirect estimation from diffusive–ebullitive relationships, increasing uncertainty in ebullition. - Highly modified systems (point-source dominated reaches, ditches, canals) were excluded from modeling due to inadequate predictors; reservoirs/impoundments (lentic) were also excluded even though they are major emitters. - The model could not be applied in Greenland and Antarctica. - Not all author affiliations were available in the provided excerpt, and some details may be truncated in the source text.